A Link and Weight-Based Ensemble Clustering for Patient Stratification

Abstract

Owing to its capability to combine multiple base clustering into a single robust consensus clustering, the ensemble clustering technique has attracted increasing attention over recent years. Although many successful clustering methods have been proposed, there is still room for improvement in the existing approaches. In this paper, we propose a novel ensemble clustering approach called a link and weight-based ensemble clustering (LWEC). We first generate a large number of similarity-indicators based on a scaled exponential similarity kernel. Then based on the similarity-indicators, an ensemble of diversified base clusterings is constructed. Further, we reckon how difficult it is to cluster an object by constructing the co-association matrix of the base clustering. And we regard related information as weights of objects. Experimental results on 35 high-dimensional cancer gene expression benchmark datasets and TCGA datasets demonstrate the efficiency and superiority of our approach.